Smart Vehicle Space (SVS) is a new smart space adopting vehicles as environment carrier, with embedded abilities of computing, communication, and perception. This paper proposes a model of driver fatigue recognition based on Hidden Markov Model (HMM) for SVS. We selected the PERCLOS feature variables as low-level contexts for driver fatigue evaluation, and established the HMM through a large number of training sample data. Then we identified the most likely driver's hidden states (high-level contexts) from the observation sequence using the Viterbi algorithm, to remind drivers to ensure the safe driving behavior. Finally, a case study in the simulation environment is given, which confirmed that the application can identify the driver's body states with a high probability, as well as maintain the good recognition effect in case of several invalid variables in observation sequence.